Team Comparisons: Good (and Bad) in the Premier League
Twelve games have passed in the Premier League and there’s value in looking at comparative historical numbers at various points. It’s a good time to take stock because a) it’s starting to become a vaguely decent sample and b) I have a load of numbers stored at twelve games because i’ve written this before. This is the eighth season of Opta data that has been published publicly around the place and with every season that passes, the more interesting the outliers become.
One reflection worth making is that at this point last season, Leicester really weren’t signalling their rise to the top. Sure they had won seven games, sat third, had scored 25 goals and lost just once, but they had also shipped 20 goals. All their shot rates were around par, and understandably given their matches averaging nearly four goals per game, their conversions were high at both ends. That they turned this into 25 points was remarkable and just stage one of the Leicester miracle; indeed all this looked sure to regress over time and see them fall back. Part two of the Leicester miracle indeed saw everything slow down, but the defence stopped conceding, almost entirely for a long spring period, the rest of the league huffed, puffed and blew itself up, and the rest is history.
I wrote a couple of weeks ago about their early accuracy problems and that this was just one of the small likely unrepeatable aspects that were likely to regress for Leicester. The early numbers show the dial has moved enormously. Between skill and luck, somewhere Leicester have stopped being able to prevent shots landing on goal and have forgotten how to create them themselves.
Their shot numbers were never great last season, but are decidedly poor this year: 10.5 shots per game (137th/160) would make Louis van Gaal proud and 2.8 shots on target per game (148th) would gain favourable glances from Tony Pulis. Over five shots on target conceded is pretty lousy too, and they will need to refocus on the league now their Champions League progression has been secured. Lightning has indeed struck twice, but while the first strike illuminated the sky and brought light where before there was only the darkness of a late night casino, this time it has wreaked only destruction having landed square onto Jamie Vardy’s left foot. Another curio is that their twelve game goals against column reads “20” for both seasons and when they were terrible in 2014/15 it was 18. Half a story, sure, but there we go.
“But what of expected goals?” Fret not, there are many ways to look at this but they’re about five goals behind expectation this season having been about twelve ahead of it in the whole of 2015-16. The inputs tell us more, they’ve lost shots and the ones they have been taking have been from an average of around a metre further away.
A kindergarten basic chart makes a point here so here you go:
There are fewer shots than there used to be in the near past. It's a point I regularly make, but it makes sense to bear in mind when pondering this stuff. We now have 160 twelve game team season starts (2009-10 to date) to look at so that's how the ranks work when mentioned.
Everything shots related looks nice for Liverpool. They lead the league for shots with 18.8 per game (9th/160), on target (7.3 per game, 4th), shots against (7.7 per game, 1st) and are in the top 20 for shots on target against too. This plays into average shots of +11.1 per game (2nd) and +4.2 on target (3rd). Of course all these things interrelate, but it's nonetheless pretty impressive stuff and they are deservedly well in the mix for the title. Comparison to 2013-14 is inevitable, but at twelve games then the team was looking good but not great--they were taking around 55% of the shots instead of the 70% they are now--and that title assault was based primarily on a huge middle and latter third where the team absolutely took off on the front end. They've already done that this season and you probably don't like money if you're betting against them based on shooting metrics.
While shot volumes like Liverpool a lot more in 2016-17 to 2013-14, there are similarities, ones that feel inevitable. At both ends in both seasons, the shots that are taken and conceded are landing on target at a high volume; this season in attack it's 39% (10th), so good, but in defence it's a huge 41% (2nd worst), while a similarly huge 15% of those shots are going in (6th worst). Weighing this up is tricky because as noted, they are conceding a tiny volume of shots but in terms of preventing the shots they allow to land on target, only Jose Mourinho's Chelsea debacle of 2015-16 exceeds this rate at this point. That sounds scary right? Well, if you're a Liverpool fan hang onto the huge excess in shots, and hope that the the defence doesn't get called to step up at any point. That is of course the key aspect here because if the attack goes through a quiet spell, which it surely will, they might need a few breaks to stay with the pace. The two good Liverpool teams are only a win apart (7-3-2 in 2013-14 to 8-3-1 now) and that lack of divergence given the differences in structural metrics could well be attributed to the 12 game save percentages, this season it is 63%, then it was 78% and only fell later on. When you commit to attack, or rabid counterpressing, there's always a possibility that there will be leaks elsewhere, but it's surely a lot more fun this way? For all that, if you could somehow glue the front third of 2016-17 to the back two thirds of 2013-14, you'd have one hell of a season.
Is this a surprise though? Well not really. Season one of Jurgen Klopp involved some excellent shot metrics and then got a little sidetracked by the Europa League run and occasional weird defeats. The potential was there and they have clearly taken a step forward this season with no Europe and a strong starting eleven. I still feel cautious about their ability to field injuries and squad depth, even allowing for fewer games but little of all this will be news to Liverpool fans, who are well served by stat types among their fanbase. They know they're good and are just hoping the train stays on the track.
Expected goals? Yep, they like them. A lot. They have a +16 goal difference because they are a expected +16 team.
Would finishing top four be enough? As a pragmatic football supporter well able to alienate swathes of fans with sensible takes, building into top four in Klopp's second season would be an excellent achievement and would bode extremely well for the future. Should a title bid fail, it should not be mourned or considered a missed opportunity, unless maybe you've sat on the Kop for 40 years, because then i'm the last person you're going to listen to.
Other big shots
In isolation, it would be easy to get excited about Liverpool's apparent progress and forget the rest but this year, a polar opposite to last, finds contenders lining up. On the front end, Tottenham, Manchester City, Chelsea, and Manchester United, yes... Manchester United are all creating historically high shot volume without doing anything spectacular, but the intrigue appears when we look at defences.
Shots Against, per game
Manchester City, 8.5 (5th/160)
Chelsea, 8.6 (6th)
Manchester United, 9.7 (9th)
We also find Tottenham, Arsenal and Southampton in the historic top twenty. So after twelve games, seven of the meanest twenty shot teams over the last eight years are from 2016-17 and this naturally feeds into on target rates too, with Chelsea's 2.2 per game (2nd) and Southampton's 2.5 (4th) particularly notable and Tottenham and Man City pretty close behind. And if we follow this wee journey to it's natural conclusion, we get various combinations of these all teams in a historic top twenty for plus/minus shots and shots on target and good old TSR and SOTR. This season, the good teams are solidly superior (and the bad teams are awful), as this example chart of TSR shows (similar holds across SOTR or xG...):
Any thoughts on a rise of the middle class are long gone here, and the bounce back from the better teams has been extreme. The "no Europe" theories around Liverpool and Chelsea must count for something, as it has freed up their schedules but Man Utd once more have an attack under Mourinho after ending up a ten shot team under van Gaal. Also, Pep Guardiola has maintained Man City's strong structure in the numbers (lest we forget this time last year they were looking superb on the numbers). Claude Puel has created an odd mix of solid volume but weird conversions at Southampton, Mauricio Pochettino is as ever the same at Tottenham and Arsenal appear to be continuing their quest towards quality over volume.
The current top five, and even Manchester United are all staring at potential top four chances, and while this far out it's too soon to be making concrete predictions of how it will shape up, 12 games in, (just about) six still vie to become four, and we are looking at a genuine deep and exciting title run.
So wait, what was that just there? Tottenham are the same as ever? What does that mean?
Slew of injuries to key men. Odd transfers in the summer. Trouble creating chances. Champions League failure. All off the back of last season's impressive work. It's a tough life being unbeaten in the league.
So what gives? All is similar, at least according to this raft of old-school numbers and also in a bigger sample, given how broadly comparable this first twelve games are to the entirety of 2015-16:
We can also look quickly at expected goals, and Tottenham are at around +0.49 per game this year compared to +0.54 last year. Recall, expected goals did not translate Tottenham's shot volumes happily then and disregards them similarly now, primarily because they shoot from anywhere and everywhere and are not focused on building better, closer chances. That's still the plan, but this year they are back towards normal in their accuracy rate, for all that in reality, it has cost them just two goals. It's sacrilege to say that aspects of this team--or at least the way it creates some of its outputs--are intriguingly similar to those that Andre Villas Boas devised, so I will not make that comparison.
One of the refreshing things about how statistical and analytical ideas are starting to spread and become understood is that I feel i've seen widespread dismissal in my echo chamber of Burnley and Hull as Premier League quality teams, despite a slew of good early results. Swansea sit rock bottom with just six points, primarily due to moderate-to-bad numbers and horrible conversion rates, and do face a genuine battle to recover but with Sunderland bad but slightly odd and the aforementioned pair in the league, must retain hope. Expected goals has them 17th with an attack-free nine shot per game Middlesbrough in hailing distance, and it seems likely that it's three from five for the drop provided Alan Pardew doesn't contrive to continue his amazingly poor run of 2016 results.
Let's line this up: Hull and Burnley are objectively and historically terrible Premier League teams and Sunderland are not far behind them.
Hull and Burnley are conceding more than twenty shots a game which ranks them worst and second worst in the sample, and the only two teams to exceed that mark.. Sunderland are conceding 18.8 per game eighth worst and each of these three is allowing over six shots on target per game and lands among the ten worst teams out of 160, with Burnley the single worst (6.8 per game).
Burnley are creating shots at a rate of 8.7 per game with 2.3 on target, enough for second and third worst rankings while Hull and Sunderland also inhabit various bottom five, ten or twenty positions.
These three teams are in the bottom four for the entire sample for all combination shot metrics (+/-, TSR, SOTR) with Burnley's shot on target ratio a truly staggering 25% and a full five percent lower than any other team. Remember the bad teams? Cardiff 2013-14, Portsmouth 2009-10, Reading 2012-13 or Aston Villa 2015-16... all of whom put up better numbers than this sorry lot at this stage of the season. Poor Tom Heaton is making more saves than any other goalkeeper has this decade. He has rightly earned plaudits but has faced no choice.
Expected goals is happy to join in and while the only team to go a season with more than a minus one per game expected goal difference is Reading 2012-13, both Burnley and Hull are exceeding this mark. Yet by virtue of the Heaton heroics and a positive skew at both ends--at least until they were trounced 4-0 at West Brom--Burnley are sitting in a hollow and fragile 12th, almost certainly doomed to fall, with only those points already accrued to use as a parachute and potentially spare their demise. Huge improvement is required to stop them falling towards the trap door, as it is for Hull who simply seem overmatched and ill prepared for the league. To at least some degree, the good teams seem better this year and have happily beaten up the bad teams, and this is reflected here.
Sunderland defy explanation year after year by surviving with terrible metrics. What more can be said?
- The difference between West Ham's rate of of getting their shots on target (20%) compared to their opposition (40%) is five percentage points higher than another team in the sample. Aston Villa's gutter ball streak 2015-16 is the worst full season here (-9%) so to be over twice as bad at 12 games is both horrific and surely prime for reversion.
- There are no historically high or low save percentages thus far.
- Our old friend PDO, still a cracking ready reckoner at this stage, sees Southampton 3rd worst after 12 at 0.79, and Swansea at 6th at 0.84. Arsenal and Burnley are both at 1.16 and high. These numbers may or may not move towards 1.00.
- Everton's rate of landing their shots on target compared to their opposition is extremely high (+12 percentage points, =2nd). The two teams ranked first here and tied second are Southampton 2014-15 and 2015-16. So: of 160 twelve game season starts, the three highest scores by this metric are the only three Ronald Koeman teams in the sample. This would be a lot more powerful if it held through the season, but it hasn't and Southampton regressed in both seasons here. But it's nonetheless the most curious of the curios.
Thanks for reading